Messages in this thread | | | Date | Thu, 18 Jul 2013 11:20:36 +0900 | From | Masami Hiramatsu <> | Subject | Re: [RFC PATCH V2] tracing: Check f_dentry before accessing event_file/call in inode->i_private |
| |
(2013/07/17 23:51), Oleg Nesterov wrote: > On 07/17, Masami Hiramatsu wrote: >> >> (2013/07/16 3:16), Oleg Nesterov wrote: >>> On 07/09, Masami Hiramatsu wrote: >>>> >>>> To avoid this, when opening events/*/*/enable, we have to ensure >>>> the dentry of the file is not unlinked yet, under event_mutex >>>> is locked. >>> >>> Probably this can work, but I am starting to think that this ref >>> count becomes toooooo complex.... >> >> Oleg, I agree that is a bit complex way to fix but it can fix. >> >> However, I think we should fix bugs first, because there are bugs. >> I don't like to leave a bug in the kernel as it is, while we re- >> start discussing new approach... > > And I am trying to fix the same problems, just _I think_ there is > another and simpler approach.
I see. I also just feel uncomfortable with leaving obvious bugs, like as having ants in my pants :)
>> So, I'd like to suggest you getting these fixes into ftrace as >> first-aid treatment, and discuss your new series. > > Well, perhaps you are right... But this TRACE_EVENT_FL_REF_MASK code > is new too, it is not that we only need a small fixlets to finish it.
Would you mean that TRACE_EVENT_FL_REF_MASK may also have some problems?
> So I think that it makes sense to discuss the alternatives before we > decide what exactly we should do.
Your approach is also interesting for me, indeed. However, it is so different from current one. I think you should clarify what bug you would like to solve and how. And we should clarify what ants^H^H^H^H bugs we have found in ftrace too. ;)
Thank you,
-- Masami HIRAMATSU IT Management Research Dept. Linux Technology Center Hitachi, Ltd., Yokohama Research Laboratory E-mail: masami.hiramatsu.pt@hitachi.com
| |